Voice: personal MCP tools + standalone assistant mode

CLAUDE.md

@@ -252,7 +252,9 @@ The voice stack is designed to be droppable into another app. Layering, strictly
 
 - **`AgentHubVoice` package** (`app/modules/AgentHubVoice`) — app-agnostic voice infrastructure: `RealtimeVoiceEngine`, audio capture/playback, dictation/STT, `VoiceTool`/`VoiceToolRegistry`, realtime session config builder, OpenAI key handling, `VoiceScreenCaptureService` (+ `VoiceScreenCaptureTools` factory for `list_displays`/`capture_screen`), and the HUD contract (`VoiceHUDHost`, `VoiceHUDConfiguration`, `VoiceHUDPresenting`). Depends only on SwiftOpenAI. Never import AgentHub types here.
 - **`AgentHubVoicePanel` target** (same package) — reusable SwiftUI HUD + onboarding (`VoiceOnboardingView`, `VoiceOption`) + `AppKitVoiceHUDPresenter`. Depends only on `AgentHubVoice`; branding and UserDefaults keys arrive via `VoiceHUDConfiguration` — never hardcode an app's name or preference keys in this target.
-- **`AgentHubCore`** — the AgentHub-specific glue: `VoiceAgentToolExecutor`, `VoiceToolCatalog` (session tools; appends the capture tools from the factory), target resolver, completion watcher, transcript reader, `VoiceControlCoordinator`, settings UI, and `AgentHubVoiceHUDHost` (the `VoiceHUDHost` adapter) plus `VoiceHUDConfiguration.agentHub`. Core imports both voice products; the voice package must never import core.
+- **`AgentHubCore`** — the AgentHub-specific glue: `VoiceAgentToolExecutor`, `VoiceToolCatalog` (session tools; appends the capture tools from the factory and the MCP tools from `VoiceMCPToolProvider`), target resolver, completion watcher, transcript reader, `VoiceControlCoordinator`, settings UI, and `AgentHubVoiceHUDHost` (the `VoiceHUDHost` adapter) plus `VoiceHUDConfiguration.agentHub`. Core imports both voice products; the voice package must never import core.
+- **Voice MCP tools** — `VoiceMCPToolProvider` bridges the user's personal MCP servers (Claude `~/.claude.json` + Codex `~/.codex/config.toml`, merged by name, Claude wins) into voice conversations as namespaced `{server}__{tool}` function tools, proxied through a dedicated `MCPAppDiscoveryService` instance with a longer request timeout (`AgentHubProvider.voiceMCPDiscoveryService`). Servers are opt-in per name via `AgentHubDefaults.voiceMCPEnabledServers` (Voice settings checklist, default off) — only enabled servers are ever spawned/contacted; MCP configs may carry secrets in `env` and must never be logged. Tool handlers have their own deadline and truncate output so a slow server can't freeze the mic-muted conversation.
+- **Voice assistant mode** — the HUD target chip's "Assistant · no session" option (`AgentHubDefaults.voiceAssistantMode`, mirrored by `VoiceHUDSettingsKeys.assistantMode`) runs conversations as a standalone assistant: no session snapshot, and the registry (`VoiceToolCatalog.makeTools(assistantMode:)`) hard-excludes every session-mutating tool (`send_prompt`, `launch_session`, `create_worktree_tasks`, `approve_pending_tool`) and screen capture, keeping read-only session tools + MCP tools. `RealtimeSessionConfigurationBuilder` keys the persona on `send_prompt` presence — registries without it get the answer-directly assistant persona. Toggling the chip mid-conversation stops the engine so a live session can't keep its old tool scope.
 
 To reuse voice in another app: depend on the `AgentHubVoice` package, implement `VoiceHUDHost`, supply a `VoiceHUDConfiguration`, and present `AppKitVoiceHUDPresenter` — plus your own tool catalog for app-specific tools.
 

app/modules/AgentHubCore/Sources/AgentHub/Configuration/AgentHubDefaults.swift

@@ -218,6 +218,16 @@ public enum AgentHubDefaults {
   /// visualizer. Type: Bool (default: false)
   public static let voiceShowTranscript = "\(keyPrefix)voice.showTranscript"
 
+  /// MCP servers (by name, from the user's Claude/Codex configs) whose tools
+  /// are exposed to voice conversations. Empty means no MCP tools.
+  /// Type: [String] (default: empty)
+  public static let voiceMCPEnabledServers = "\(keyPrefix)voice.mcpEnabledServers"
+
+  /// Whether voice conversations run as a standalone assistant: no session
+  /// target, no session snapshot, and no tools that push prompts or content
+  /// into a session. Type: Bool (default: false)
+  public static let voiceAssistantMode = "\(keyPrefix)voice.assistantMode"
+
   // MARK: - Feature Flags
 
   /// Whether smart mode (AI-powered orchestration planning) is enabled

app/modules/AgentHubCore/Sources/AgentHub/Configuration/AgentHubProvider.swift

@@ -270,9 +270,19 @@ public final class AgentHubProvider {
     targetResolver: voiceSessionTargetResolver
   )
 
+  /// Dedicated MCP gateway for voice tool calls. Separate instance from the
+  /// MCP Apps discovery service so long-running tool calls get a longer
+  /// request timeout without affecting app-resource discovery.
+  public private(set) lazy var voiceMCPDiscoveryService: any MCPAppDiscoveryServiceProtocol =
+    MCPAppDiscoveryService(requestTimeoutSeconds: 45)
+
+  public private(set) lazy var voiceMCPToolProvider: any VoiceMCPToolProviding =
+    VoiceMCPToolProvider(discovery: voiceMCPDiscoveryService)
+
   public private(set) lazy var voiceToolCatalog: any VoiceToolCataloging =
     VoiceToolCatalog(
       executor: voiceToolExecutor,
+      mcpToolProvider: voiceMCPToolProvider,
       onBackgroundWaitCountChanged: { [weak self] count in
         self?.realtimeVoiceEngine.setAwaitingBackgroundWork(count > 0)
       }
@@ -532,10 +542,15 @@ public final class AgentHubProvider {
   }
 
   public func shutdownMCPAppDiscoveryService() {
-    let service = mcpAppDiscoveryService
+    let services: [any MCPAppDiscoveryServiceProtocol] = [
+      mcpAppDiscoveryService,
+      voiceMCPDiscoveryService,
+    ]
     let semaphore = DispatchSemaphore(value: 0)
     Task.detached(priority: .utility) {
-      await service.shutdown()
+      for service in services {
+        await service.shutdown()
+      }
       semaphore.signal()
     }
     _ = semaphore.wait(timeout: .now() + 2.0)

app/modules/AgentHubCore/Sources/AgentHub/Services/AgentHubVoiceHUDHost.swift

@@ -28,7 +28,8 @@ extension VoiceHUDConfiguration {
         hudFrame: AgentHubDefaults.voiceHUDFrame,
         screenCaptureEnabled: AgentHubDefaults.voiceScreenCaptureEnabled,
         onboardingCompleted: AgentHubDefaults.voiceOnboardingCompleted,
-        showTranscript: AgentHubDefaults.voiceShowTranscript
+        showTranscript: AgentHubDefaults.voiceShowTranscript,
+        assistantMode: AgentHubDefaults.voiceAssistantMode
       ),
       accentColor: .brandSecondary
     )
@@ -62,14 +63,20 @@ public final class AgentHubVoiceHUDHost: VoiceHUDHost {
       .map(Self.hudTarget)
   }
 
+  private var isAssistantMode: Bool {
+    defaults.bool(forKey: AgentHubDefaults.voiceAssistantMode)
+  }
+
   public func makeToolRegistry() -> VoiceToolRegistry {
     VoiceToolRegistry(
-      tools: provider?.voiceToolCatalog.makeTools() ?? []
+      tools: provider?.voiceToolCatalog.makeTools(
+        assistantMode: isAssistantMode
+      ) ?? []
     )
   }
 
   public func makeSessionContext() -> String? {
-    guard let provider else { return nil }
+    guard let provider, !isAssistantMode else { return nil }
     return VoiceSessionContextBuilder.make(
       summary: provider.voiceToolExecutor.listSessions()
     )

app/modules/AgentHubCore/Sources/AgentHub/Services/VoiceControlCoordinator.swift

@@ -15,6 +15,7 @@ public final class VoiceControlCoordinator {
   private let presenter: any VoiceHUDPresenting
   private let defaults: UserDefaults
   private let hotKey: GlobalHotKey
+  private let onHUDShown: (@MainActor () -> Void)?
   private var isStarted = false
 
   public var isHUDVisible: Bool {
@@ -25,12 +26,14 @@ public final class VoiceControlCoordinator {
     registrar: any GlobalHotKeyRegistrarProtocol,
     presenter: any VoiceHUDPresenting,
     defaults: UserDefaults = .standard,
-    hotKey: GlobalHotKey = .voiceHUDDefault
+    hotKey: GlobalHotKey = .voiceHUDDefault,
+    onHUDShown: (@MainActor () -> Void)? = nil
   ) {
     self.registrar = registrar
     self.presenter = presenter
     self.defaults = defaults
     self.hotKey = hotKey
+    self.onHUDShown = onHUDShown
   }
 
   public convenience init(
@@ -40,7 +43,12 @@ public final class VoiceControlCoordinator {
     self.init(
       registrar: CarbonGlobalHotKeyRegistrar(),
       presenter: provider.makeVoiceHUDPresenter(defaults: defaults),
-      defaults: defaults
+      defaults: defaults,
+      onHUDShown: { [weak provider] in
+        // Warm the MCP tool cache so the conversation the user is about to
+        // start sees the enabled servers' tools.
+        provider?.voiceMCPToolProvider.scheduleRefresh()
+      }
     )
   }
 
@@ -87,6 +95,9 @@ public final class VoiceControlCoordinator {
   }
 
   public func toggleHUD() {
+    if !presenter.isVisible {
+      onHUDShown?()
+    }
     presenter.toggle()
   }
 }